Exclusive: Here's how Tata's new AI-driven mutual fund plans beat the market and how it will react if a war breaks out
Tata Quant Fundis India’s first fund that claims it can beat the market using artificial intelligence by cutting out human bias.
- Business Insider spoke to
Utpal Sharma, the Head of Business Analytics at Tata Asset Management about how the fund works and how it would react in case of unpredictable events — like wire.
- Sharma also explains why the Tata Quant Fund doesn’t use National Language Processing (NLP) and what the main challenges are in using AI for predicting the market.
In fact, there are already plenty of funds using AI in the US and Europe. India is joining the trend with
Instead of having fund managers analyse market trends to pick the best stocks, a combination of rule-based algorithms picks the winners and losers. Some, like the $54 billion Pangora Asset Management Quant Fund, even use natural language processing (NLP) to keep tabs on traders to predict upcoming market trends.
"We have tried that, we didn’t get much of a lift from using that data. So, currently, we are not reading opinions and we’re not reacting to events. We have got structured data in place and making calls based on that structured data," Utpal Sharma, the Head of Business Analytics at Tata Asset Management told Business Insider.
But, human intervention comes with its own set of advantages like experience, the ability to understand complex problems and instinct. And, Indian markets — since they are a part of a developing economy — can sometimes be unpredictable.
"In our structure, though we are machine-driven, we have a fund manager. In the event of something seriously very different from what has happened in the past years occurs — like a war — he can intervene," explained Sharma.
Why do we want to remove human judgement?
One of the primary reasons to use AI to beat the market is to remove human bias from affecting choices. But it’s not that easy to measure what needs to be left out.
"It’s very hard to pinpoint how human biases impacted what performances by quantum. That’s not something that we measured. We definitely think if you do it in a systematic manner where you are eliminating judgements which are not completely objectively, which are not non-deterministic," said Sharma.
Using AI to beat the market
Tata Mutual Funds plans to use artificial intelligence (AI) and machines learning (ML) to beat the market. It is India’s first mutual fund to offer an AI-driven investment strategy. Even though the fund may not know how to react to a situation like an unexpected war, that’s not a day-to-day occurrence.
For normal events like policy changes, budget declarations and earnings, the model is geared to predict how the market will react — and figure out how to beat it. It relies on data from the past 22 years to forecast how the market will move and they claim that they’ve been testing the model since 2011.
"Yes [it will beat the market]. So our index is the BSE 200. Our objective would be to deliver better results than the index. The index itself has generated around — from January to December — 10.25 per cent. Our objective is to do better than those returns," he said.
However, not every quant fund is a story of success. The AI Global Equity Fund managed by Aberdeen Standard Investment underperformed in comparison to the market. The explanation offered by the fund manager was that financial market data is finite and that algorithms can only learn so much from past data.
"We got two models. One which chooses the best portfolio to invest in, in the coming period. Another model which predicts whether the absolute direction is positive or negative. When we combine them, we have the ability to participate in uptrending markets while avoiding markets which are headed downwards," said Sharma.
The model makes its predictions on a monthly basis and is retrained every six months, learning from past events.
"We have tried this for different frequencies of relearning and tweaking the model based on recent data. We found that when we take six months incremental data and rebalance or readjust our models, that when we see the best results," explained Sharma.
Rise of the machines: 2020 will be a good year for India to teach AI some ethics
India wants to be on the cusp of artificial intelligence but lacks the laws to back it up
5G airports, AI-powered safe cities and many more that Huawei has on offer at the India Mobile Congress 2019
Popular on BI
- A Google employee of 11 years says he and his wife stared at each other in 'disbelief' when they realized they'd both been laid off by the company
- A Google engineer of 8 years says his 'spidey-senses' detected incoming layoffs — and felt 'isolated' when his 'faceless' severance email arrived
- 6 signs that you're in a one-sided relationship and how to find balance
- 70-plus Indian startups show exit door to 21K techies, more pink slips coming
- Steps taken in subsidence-hit Joshimath inadequate, declare Himalayas eco-sensitive zone, say experts
- An app that wants to turn fear of stock markets into fun raises ₹75 lakh from sharks
- Google may release Pixel Tablet Pro with Tensor G2 chip
- BMW launches its 3rd gen BMW X1 in India